CN113033663A - Automatic container terminal equipment health prediction method based on machine learning - Google Patents
Automatic container terminal equipment health prediction method based on machine learning Download PDFInfo
- Publication number
- CN113033663A CN113033663A CN202110324789.6A CN202110324789A CN113033663A CN 113033663 A CN113033663 A CN 113033663A CN 202110324789 A CN202110324789 A CN 202110324789A CN 113033663 A CN113033663 A CN 113033663A
- Authority
- CN
- China
- Prior art keywords
- equipment
- abnormal
- state
- data
- health
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an automatic container terminal equipment health prediction method based on machine learning. The invention provides an automatic container terminal equipment health index calculation method aiming at the current situations that the abnormal types of cooperative equipment in an ultra-large container terminal are complex and changeable and the health degree is difficult to judge, and solves the defect that the traditional method can only carry out normal or abnormal judgment on the equipment and cannot accurately evaluate the health degree of the equipment. The method adopts a principal component analysis method to carry out dimensionality reduction on high-dimensional monitoring data, extracts the data characteristics of a hidden layer through an LSTM network, designs an automatic container terminal equipment health index calculation formula according to the relation between the equipment state and the hidden layer characteristics, and further monitors the equipment health degree and the abnormal development trend.
Description
The technical field is as follows:
the invention relates to the field of anomaly detection and machine learning, in particular to a method for extracting hidden layer characteristics based on an LSTM network and evaluating the current health degree and the anomaly development trend of equipment of an automatic container terminal according to the relation between the equipment state and the hidden layer characteristics.
Background art:
under the complex scene of super large container terminal, various equipment collaborative operations, in case break down the production in a comprehensive way, disturb whole production plan, cause great economic loss for the pier enterprise, the industrial scene of modernization puts forward higher requirement to the fault detection technique, not only need can judge whether equipment is normally working, but also need can carry out the analysis to equipment health degree, and then handles the trouble that probably appears in advance. The health problem of equipment is more and more paid attention by people, most of the prior technical methods can accurately judge the normal and abnormal states of the automatic equipment of the container terminal, and a method capable of analyzing the health degree and the abnormal development trend of the equipment is lacked. The invention provides an automatic container terminal equipment health degree and abnormal development trend evaluation method based on machine learning.
Closest to the prior art and its analysis:
for the problem of the evaluation of the health degree of the device, a learner also obtains a reconstruction error through a coding-decoding model to perform the evaluation of the health state of the device (Malhotra P, TV V, Ramakrishnan a, et al. Pankaj Malhotra et al proposed a Long Short Term Memory-based Encoder-Decoder (LSTM-ED) model for multi-sensor time series data, which learns data in a normal state, obtains a data reconstruction error using an encoding-decoding structure, and determines the health of a device according to the magnitude of the error, but this method can only determine the degree of deviation between the current state and the normal state of the device, and cannot analyze the abnormal trend of the device.
Disclosure of Invention
The invention aims to provide an automatic container terminal equipment health degree and abnormal development trend evaluation method based on machine learning. By learning the normal state and the abnormal state at the same time, hidden layer characteristics of all states are acquired based on the LSTM unit, and the distance relation between the characteristic value in the current state and the characteristic values of the normal state and the abnormal state is constructed, so that the health degree of the equipment is analyzed, and the possible abnormal state trend can be judged.
The technical scheme is as follows:
an automatic container terminal equipment health prediction method based on machine learning is characterized in that: when the abnormal data set of the equipment is processed, the current fault is judged and classified into one of four systems caused by the abnormality by analyzing the characteristics of the fault in the actual production process and whether the sensor data related to the data set containing the fault exceeds the normal range, wherein the four systems are a navigation system, an electric control system, a hydraulic system and an equipment management system. And summarizing the fault types into navigation abnormity, electric control abnormity, hydraulic abnormity and equipment management system abnormity, learning the normal state and the four abnormal states by using an LSTM network, extracting the characteristics of the hidden layer data, and calculating the data center and the maximum distance from the data center in the normal state to the data center in the normal state. And for the time segment to be detected, computing the hidden layer data features through the LSTM network, and respectively computing the distances between the hidden layer data features and the centers of the hidden layer data in the normal state and the four abnormal types. And if the distance between the state to be detected and the center of the normal state is less than the maximum distance in the normal state, judging that the current state is the normal state, and simultaneously calculating a health value according to the distances from the centers of the other four abnormal centers, wherein the health value is more healthy when the distance is longer, and the trend of generating corresponding abnormal is larger when the distance is closer. Otherwise, if the state to be measured is too far away from the normal state, the state is judged to be abnormal, and the abnormal center type closest to the state to be measured is the current abnormal state.
The method comprises the following implementation processes: firstly, according to the classification and setting of health indexes of automatic guide vehicle equipment of an automatic container terminal, signals detected by sensors distributed on the equipment are collected to form a data set; then, acquiring a continuous time segment by using a sliding window mechanism, and reducing the dimension of the time sequence data of the high-dimensional equipment monitoring by using a PCA (principal component analysis) model; then, sending the data after dimensionality reduction into an LSTM network for learning, and extracting hidden layer characteristics; and finally, calculating the central points of the hidden layer features in different equipment states and the Euclidean distance between the central points and the Euclidean distance, obtaining a health index through processing operation, and evaluating the health state of the equipment.
The technical scheme of the invention is based on the phenomenon that the classification of the abnormal types of the automatic container terminal equipment has potential connection, firstly, the abnormal types are summarized according to experience knowledge, then, the PCA method is utilized to reduce the dimension of high-dimensional equipment detection data, the data after dimension reduction is sent to an LSTM network, the equipment state is learned, the hidden layer characteristics of the data time dimension are obtained, the Euclidean distance between the hidden layer characteristics of different states is calculated, the health degree of the equipment is judged according to the distance between the state to be detected and the normal state, the trend size of each abnormal occurrence is calculated according to the distance between the state to be detected and the abnormal state, and therefore, the health degree of the equipment and the development trend of the abnormal are evaluated.
By adopting the scheme, the invention has the beneficial effects that:
1. the invention provides a method for specifying the health index of equipment in an automatic container terminal, which effectively utilizes the time dimension characteristics of data to judge the health state of the equipment, thereby predicting the health trend of the equipment in a period of time in the future and achieving the effect of early warning of the equipment;
2. the method overcomes the defect that the traditional method can only judge the deviation degree of the equipment from the normal state but cannot analyze the abnormal trend of the equipment, fully utilizes the data relation among hidden layer characteristics, can judge the health degree of the equipment, can indicate the abnormal development trend of the equipment, and has great guiding significance.
Drawings
Fig. 1 is a basic flow chart of the health and abnormal development trend evaluation of the automatic container terminal equipment based on machine learning according to the invention.
Fig. 2 shows the percentage variance of the first 30 eigenvalues when PCA dimension reduction analysis is performed on training set a.
Fig. 3 is an LSTM network structure.
Fig. 4 is a schematic diagram of Automatic Guided Vehicle (AGV) exception category classification in an automated container terminal.
Fig. 5 shows hidden layer feature values in different states.
FIG. 6 is a graph showing the results of the experiment according to the present invention.
FIG. 7 is an example of health values and health assessments.
Detailed Description
The present embodiment will be further described with reference to the flowchart shown in fig. 1.
Examples
Step 1: the classification and setting of the health index of the automatic guided vehicle equipment of the automatic container terminal are as follows:
automatic Guided Vehicle (AGV) equipment of an Automatic container terminal is classified into four systems: navigation system, electronic control system, hydraulic system and equipment management system.
Wherein, navigation mainly is responsible for absolute position correction and dead reckoning, and both ends installation RFID antenna is gathered and is read ground magnetic nail around through the automobile body, obtains:
the front Antenna newly detects the magnetic pin absolute coordinates t _ Antenna _ F _ AbsX, t _ Antenna _ F _ AbsY, the rear Antenna newly detects the magnetic pin absolute coordinates t _ Antenna _ R _ AbsX, t _ Antenna _ R _ AbsY, and
coordinates of the center points of the front and rear antennas, namely Localization _ antenna _ F _ X, Localization _ antenna _ F _ Y, Localization _ antenna _ R _ X and Localization _ antenna _ R _ Y, are used for realizing absolute position positioning of the vehicle.
The magnetic nails are laid on the ground of the wharf at intervals of 2 meters, and the identity numbers and the absolute positions of the magnetic nails in a wharf coordinate system are recorded in the magnetic nails. When the front and the rear antennae of the AGV read the magnetic nails at the same time, the absolute position and the driving direction of the AGV in the field can be calculated. The antenna reads the magnetic nail, so that the absolute position calculation of the vehicle running deviation is realized, and a basis is provided for the deviation correction of the wheel.
The 4 Wheel speed encoders mounted at the axle positions correspondingly measure the rotating speeds Wheel _ FL _ AngleEncoder, Wheel _ FR _ AngleEncoder, Wheel _ RL _ AngleEncoder and Wheel _ RR _ AngleEncoder of the 4 Wheel axles, and then calculate the running speed of each Wheel.
The angle encoders installed on rotating shafts of 4 wheels correspondingly measure rotating angles Wheel _ FL _ SteerAngle, Wheel _ FR _ SteerAngle, Wheel _ RL _ SteerAngle and Wheel _ RR _ SteerAngle of the 4 wheels, and the angle deviation of two wheels of the same axle meets the Ackerman principle requirement. The combination wheel speed, the angle measurement value and the AGV body motion model can simulate and calculate the motion condition of the AGV, including the motion speed and the motion angle, the position calculation when the magnetic nails cannot be read by the AGV in a storage yard is realized, and then the running position of the AGV is calculated.
The gyroscope is mainly used for measuring the AGV Heading angle Gyro _ Heading and the acceleration Gyro _ acceleration. The movement state of the AGV can be calculated based on the angle and acceleration data of the gyroscope by combining the above ground magnetic nail positioning mode, the movement state comprises dead reckoning, and the calculation result of the gyroscope is used for verifying whether the calculation value based on the wheel speed and the angle is accurate or not, so that the accuracy of the dead reckoning is ensured. Here, the angle calculation value of the gyroscope is used as the dead reckoning use value.
The planar laser sensors are arranged on the head and the tail of the vehicle and used for scanning whether barriers exist in the front of the driving direction of the vehicle, the protection range is related to the vehicle speed (s is 7v +2), 4 protection strategies including straight driving, turning, crab driving and support driving protection strategies are adopted by means of sector scanning areas, and the protection range with the corresponding shape is made based on the size and the shape of the area to be covered when the vehicle drives. When the obstacle is scanned in the protection range, the deceleration and parking actions are generated according to the distance.
When the navigation system breaks down, data collected by sensors such as an appointed RFID antenna, a wheel speed encoder, a gyroscope, a plane laser sensor and the like can deviate from normal values, and accordingly abnormal occurrence points can be accurately found out.
The AGV is mainly responsible for managing and controlling the work of an electric system, the AGV is provided with 1 charger, 2 walking inverters and 1 hydraulic inverter, and through a built-in VOLTAGE and current transformer and a temperature sensor heat dissipation relay working state detection sensor, front and rear driving DC VOLTAGEs DC _ INV _ F _ IW10, VDC _ INV _ R _ IW10, driver DC VOLTAGEs DC _ INV _ HYD _ IW10, controller supply VOLTAGE ControlSUPPLY _ VOLTAGE, relay working state TraceDataBool _ VSDiagnose1 and other information are obtained to detect the working states of working VOLTAGE, current, temperature and a heat dissipation system.
The proximity limits offset _ toe _ F and offset _ toe _ R of the AGV crash bumper (mechanical structure, which swings when impacted) are used to detect whether an impact action occurs. The signal TraceDataBool _ tamp1 is obtained by using the jacking platform ultrasonic sensor and is used for detecting whether a barrier exists in a 75cm range of the signal TraceDataBool _ tamp1 (when the signal is jacked to a support to send a box in combination with the operating condition of the AGV, the signal works by using the sensor), and the sensor does not work when the AGV jacks. The container box type (20, double 20, 40 and 45 feet) planted by the AGV is judged by utilizing the box type limit (ultrasonic wave and photoelectric limit) of the container taking and placing guide plate arranged on the upper part of the vehicle body and the combinational logic of the signal TraceDataBool _ tamp 2.
The hydraulic system is mainly responsible for Steering, jacking and Brake control of equipment, and mainly collects main system pressure information and Brake and jacking platform state information through a pressure sensor, a temperature and humidity sensor, a normally closed Brake opening pressure sensor, a jacking platform displacement sensor and the like, such as front-middle and rear jacking magnetic scale displacement F _ L _ PAGE2_ Steering _ Ref, F _ R _ PAGE2_ Steering _ Ref, R _ L _ PAGE2_ Steering _ Ref, R _ R _ PAGE2_ Steering _ Ref, PAGE12_ Brake _ FrtWheel, PAGE12_ Brake _ BakWheel, front-rear axle pressure feedback VS _ BigPurpop 05_1, VS _ Smallmp _ PV05_2, Steering proportional valve feedback VS _ Steer, hydraulic driver speed and power feedback VS _ HyaudredSpeedDrivePowerk, VS _ Hyaudrwed. When a hydraulic system breaks down, data collected by a specified pressure sensor, a temperature and humidity sensor, a normally closed brake opening pressure sensor, a jacking platform displacement sensor and the like can show a special rule, and the specific faults of the hydraulic system can be analyzed according to different phenomena of different sensors.
The equipment management system mainly acquires fault state information through command verification according to motion working conditions, such as a USINT command type Vms2NSCmd _ TCP _ Commandtype, a gyroscope correction state TraceDataBool _ tamp6, a power-taking trolley stretching state TraceDataBool _ tamp3, a DC voltage working state TraceDataBool _ VSdiagnosese and the like. Different combinations of data represent different device management system exceptions.
The signals detected by the sensors described above constitute a data set.
An Automatic Guided Vehicle (AGV) data set of an automatic container terminal is randomly divided into an original training set and an original testing set according to a ratio of 9: 1. Various abnormal types in the data set are divided into navigation abnormality, electric control abnormality, hydraulic abnormality and equipment management system abnormality as shown in fig. 4, and a training set A and a test set B are generated.
The data set used in the experiment of this example was 250 sensor monitoring data collected on-site from an Automated Guided Vehicle (AGV) at an automated container terminal at a Qingdao harbor. The data set consisted of 21 normal data sets and 91 abnormal data sets, each data set having a size of about 180000 × 250, and a sampling frequency of 92 times per second.
Step 2: training a PCA model by using a data set A, reducing the dimension of the monitoring data of an Automatic Guided Vehicle (AGV) sensor of an automatic container terminal, and carrying out normalization processing, wherein the variance percentage of more than 90% of the original data can be reserved by the first 30 characteristic values as shown in FIG. 2;
and step 3: sampling every 30 times to obtain a time slice, splitting the dimensionality-reduced data set A by using a sliding window mechanism, training an LSTM network comprising three LSTM layers and a full connection layer by using the split data, wherein the optimization algorithm uses an RMSProp algorithm, the loss function uses a cross entropy loss function, and the network structure is shown in FIG. 3;
and 4, step 4: different from the traditional LSTM method which directly carries out full connection on the last layer of LSTM hidden layer features to obtain a classification result, the feature values of the last layer of LSTM unit nodes are extracted to obtain the sample hidden layer feature values of a normal state, a navigation abnormality, an electric control abnormality, a hydraulic abnormality and an equipment management system abnormality, and the feature values are used for training a new PCA model to reduce the obtained feature values to two dimensions, so that the hidden layer features of each state have an obvious distance relation. As shown in fig. 5: in the figure, the value 1 represents a normal state, and the values 2, 3, 4 and 5 represent an electric control abnormality, a hydraulic abnormality, a navigation abnormality and an equipment management system abnormality respectively. (ii) a
And 5: respectively calculating the characteristic value central point of each state obtained in the step 4 by using a formula (1)
Wherein k represents the number of each state time segment in the data set, and x and y represent sample characteristic values extracted by the LSTM;
step 6: sampling the data set B to be tested into a time segment every 30 times, collecting the time segment according to a sliding window mechanism, and normalizing the data by using the mean value and the variance of the training set A;
and 7: using the PCA model trained in the step 2 to reduce the dimension of the data B to be tested, and providing a step 8;
and 8: sending the data to be detected into the LSTM network trained in the step 3, acquiring hidden layer characteristics, and providing the hidden layer characteristics to the step 9;
and step 9: reducing the dimension of the hidden layer feature obtained in the step 8 by using the new PCA model obtained in the step 4 to obtain a new feature value (x, y), then calculating the Euclidean distance with the abnormal state center value obtained in the step 5, and obtaining a sample health index H and an abnormal trend i through a formula (2) for providing to the step 10;
wherein (x)0,y0) Representing the center point of the characteristic value in the normal state in the data set, (x)i,yi) Substitute for Chinese traditional medicineThe characteristic value central point, dis (a, b) in the abnormal state in the table data set represents the Euclidean distance between the values a and b;
step 10: and performing health evaluation on the current equipment according to the obtained health factor H, and indicating the most probable direction of the abnormality according to the value of i, namely the ith abnormality is the most probable abnormality.
The invention mainly provides a method for calculating the health index of equipment of an automatic container terminal, and therefore, the invention also makes an additional comparison test, utilizes an LSTM-ED algorithm to train and test an original data set, the result is shown in figure 6, the test result shows that the LSTM-ED algorithm can not accurately detect abnormal states, and the health state can be effectively predicted on the basis of an LSTM network by the method, the accuracy rate reaches 90.6%, and in addition, as shown in figure 7, the health value not only can judge the health state of the equipment, but also can effectively predict the abnormal development trend of the equipment.
Claims (2)
1. An automatic container terminal equipment health prediction method based on machine learning is characterized in that: when an equipment abnormal data set is processed, judging and classifying the current fault into one of four systems caused by abnormality by analyzing the characteristics of the fault in the actual production process and whether the sensor data related to the data set containing the fault exceeds a normal range, wherein the four systems are a navigation system, an electric control system, a hydraulic system and an equipment management system; the fault types are summarized into navigation abnormity, electric control abnormity, hydraulic abnormity and equipment management system abnormity, the LSTM network is utilized to learn the normal state and the four abnormal states, the characteristics of the hidden layer data are extracted, and the data center and the maximum distance from the data center in the normal state to the data center in the normal state are calculated; for the time segment to be detected, calculating the hidden layer data characteristics through the LSTM network, and respectively calculating the distances between the hidden layer data characteristics and the centers of the hidden layer data in a normal state and four abnormal types; if the distance between the state to be detected and the center of the normal state is smaller than the maximum distance in the normal state, judging that the current state is the normal state, and simultaneously calculating a health value according to the distances from the other four abnormal centers, wherein the more the distance is, the healthier the more the distance is, and the more the distance is, the larger the trend of corresponding abnormality is; otherwise, if the state to be measured is too far away from the normal state, the state is judged to be abnormal, and the abnormal center type closest to the state to be measured is the current abnormal state.
2. The automated container terminal equipment health prognosis method based on machine learning of claim 1, wherein: the realization process is as follows: firstly, according to the classification and setting of health indexes of automatic guide vehicle equipment of an automatic container terminal, signals detected by sensors distributed on the equipment are collected to form a data set; then, acquiring a continuous time segment by using a sliding window mechanism, and reducing the dimension of the time sequence data of the high-dimensional equipment monitoring by using a PCA (principal component analysis) model; then, sending the data after dimensionality reduction into an LSTM network for learning, and extracting hidden layer characteristics; and finally, calculating the central points of the hidden layer features in different equipment states and the Euclidean distance between the central points and the Euclidean distance, obtaining a health index through processing operation, and evaluating the health state of the equipment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110324789.6A CN113033663A (en) | 2021-03-26 | 2021-03-26 | Automatic container terminal equipment health prediction method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110324789.6A CN113033663A (en) | 2021-03-26 | 2021-03-26 | Automatic container terminal equipment health prediction method based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113033663A true CN113033663A (en) | 2021-06-25 |
Family
ID=76474104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110324789.6A Pending CN113033663A (en) | 2021-03-26 | 2021-03-26 | Automatic container terminal equipment health prediction method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113033663A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109238735A (en) * | 2018-08-08 | 2019-01-18 | 上海交通大学 | The malfunction monitoring diagnostic system of the electronic AGV of port cargo |
| CN109446187A (en) * | 2018-10-16 | 2019-03-08 | 浙江大学 | Complex equipment health status monitoring method based on attention mechanism and neural network |
| CN110441065A (en) * | 2019-07-04 | 2019-11-12 | 杭州华电江东热电有限公司 | Gas turbine online test method and device based on LSTM |
| CN110826335A (en) * | 2019-11-14 | 2020-02-21 | 北京明略软件***有限公司 | Named entity identification method and device |
| CN111368890A (en) * | 2020-02-26 | 2020-07-03 | 珠海格力电器股份有限公司 | Fault detection method and device and information physical fusion system |
| CN111461551A (en) * | 2020-04-01 | 2020-07-28 | 未必然数据科技(北京)有限公司 | Electric submersible pump fault early warning method based on deep learning and SPC (statistical computer) criterion |
| CN111813084A (en) * | 2020-07-10 | 2020-10-23 | 重庆大学 | Mechanical equipment fault diagnosis method based on deep learning |
-
2021
- 2021-03-26 CN CN202110324789.6A patent/CN113033663A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109238735A (en) * | 2018-08-08 | 2019-01-18 | 上海交通大学 | The malfunction monitoring diagnostic system of the electronic AGV of port cargo |
| CN109446187A (en) * | 2018-10-16 | 2019-03-08 | 浙江大学 | Complex equipment health status monitoring method based on attention mechanism and neural network |
| CN110441065A (en) * | 2019-07-04 | 2019-11-12 | 杭州华电江东热电有限公司 | Gas turbine online test method and device based on LSTM |
| CN110826335A (en) * | 2019-11-14 | 2020-02-21 | 北京明略软件***有限公司 | Named entity identification method and device |
| CN111368890A (en) * | 2020-02-26 | 2020-07-03 | 珠海格力电器股份有限公司 | Fault detection method and device and information physical fusion system |
| CN111461551A (en) * | 2020-04-01 | 2020-07-28 | 未必然数据科技(北京)有限公司 | Electric submersible pump fault early warning method based on deep learning and SPC (statistical computer) criterion |
| CN111813084A (en) * | 2020-07-10 | 2020-10-23 | 重庆大学 | Mechanical equipment fault diagnosis method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102629298B (en) | A kind of Rail Transit System security of operation appraisal procedure | |
| CN107010053A (en) | Method for selecting optimization track | |
| CN109649396B (en) | Safety detection method for commercial vehicle driver | |
| CN103471865A (en) | Train suspension system failure isolation method based on LDA method | |
| CN109917213B (en) | Contact network detection fault early warning method based on dimensionality reduction fusion and factor analysis | |
| CN108710637A (en) | Taxi exception track real-time detection method based on time-space relationship | |
| CN115824313A (en) | Integrated multi-parameter monitoring control method and system for grain condition monitoring | |
| CN113435404B (en) | Electric bus shunting safety auxiliary driving method and system based on radar and image detection | |
| CN113295212B (en) | Landslide overall stability safety early warning system applying multi-monitoring point synergistic effect | |
| Ghiasi et al. | An unsupervised anomaly detection framework for onboard monitoring of railway track geometrical defects using one-class support vector machine | |
| CN117141472A (en) | Active obstacle avoidance control system and method for vehicle running | |
| CN112015180B (en) | Intelligent experiment trolley and control system thereof | |
| CN113033663A (en) | Automatic container terminal equipment health prediction method based on machine learning | |
| CN109633657A (en) | A kind of method and system that barrier is tracked based on multiple ultrasonic radars | |
| CN109406076A (en) | A method of beam bridge structure damage reason location is carried out using the mobile principal component of displacement sensor array output | |
| CN203580808U (en) | Detector for carrying out real-time detection on vehicle driving on wrong lane in simulated tunnel test | |
| CN104271428B (en) | Method for investigating Wheel Rail Contact | |
| CN210591925U (en) | Axle box vibrating wheel diameter difference detection system | |
| Gordon et al. | Vibration transmission from road surface features: Vehicle measurement and detection | |
| CN113654633B (en) | Metering detection method of dynamic truck scale weighing system under abnormal driving behavior condition | |
| Gräbe et al. | Advancing the digital railway with 4IR technologies | |
| CN110450817B (en) | Axle box vibration wheel diameter difference detection system and method | |
| CN114595514B (en) | Intelligent track safety calculation method | |
| Radicioni et al. | Overcoming strain gauges limitation in the estimation of train load passing on a bridge through deep learning | |
| CN116691768B (en) | Switch monitoring method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210625 |
|
| WD01 | Invention patent application deemed withdrawn after publication |